def solve(self, filename):
"""implements the Simulated Annealing algorithm for VRPTW"""
self.problem_file = filename
if self.trace:
global fig
plt.ion()
fig = plt.figure()
fig.canvas.set_window_title('Simulated Annealing')
else:
self.histograms = False
self.problem = Problem()
self.problem.load_problem(filename)
#problem.eliminate_time_windows()
start_time = time.time()
self.current = random_solution(self.problem)
#current = insertion1_vrptw(problem)
self.run()
self.best.endtime = time.time()
self.best.starttime = start_time
if self.trace:
self.best.plot(True)
self.present_statistics()
plt.pause(0.0001)
input("Push <ENTER> to finish")
plt.ioff()
def solve_vrptw(problem, heuristic, max_iter, display=True):
"""implements the Stochastic Hill Climbing algorithm for VRPTW"""
start_time = time.time()
current = random_solution(problem)
#current = insertion1_vrptw(problem)
ini_sol = current
if display:
current.plot()
fig.canvas.draw()
last_improved = 0
print("heuristic:", heuristic.name)
advance = max_iter // 10
it = 0
for i in range(max_iter):
it += 1
candidate = create_neighbor(current, heuristic, current.cost())
if candidate.cost() < current.cost():
current = candidate
last_improved = i+1
if display:
current.plot()
fig.canvas.draw()
if display:
print(" > iteration=%d, curr= %g" % (i+1, current.cost()))
elif it == advance:
it = 0
print("%d%%" % (int(((i+1) / max_iter) * 100)))
current.endtime = time.time()
current.starttime = start_time
if display:
current.plot(True)
fig.canvas.draw()
perturbative_statistics()
PerturbativeHeuristic.plot()
return ini_sol, current, last_improved
def simulated_annealing(filename, max_temp, min_temp, eq_iter, temp_change,
trace=True, histograms=False):
"""implements the Simulated Annealing algorithm for VRPTW"""
if trace:
global fig
plt.ion()
fig = plt.figure()
fig.canvas.set_window_title('Simulated Annealing')
else: histograms = False
problem = Problem()
problem.load_problem(filename)
#problem.eliminate_time_windows()
start_time = time.time()
current = random_solution(problem)
#current = insertion1_vrptw(problem)
temp = max_temp
best = copy.copy(current)
if trace:
best.plot()
fig.canvas.draw()
plt.pause(0.0001)
i = 0
perturbative_heuristics()
while temp > min_temp:
eiter = 0
while eiter < eq_iter:
i += 1
candidate = create_neighbor(current, best.cost(), histograms)
if should_accept(candidate, current, temp): current = candidate
if candidate.cost() < best.cost():
best = copy.copy(candidate)
if trace:
best.plot()
fig.canvas.draw()
plt.pause(0.0001)
if trace: print(" > iteration=%d, temp=%g, curr= %g, best=%g" %
(i,temp,candidate.cost(), best.cost()))
eiter += 1
temp *= temp_change
best.endtime = time.time()
best.starttime = start_time
if trace:
best.plot(True)
plt.pause(0.0001)
perturbative_statistics()
if histograms:
PerturbativeHeuristic.plot()
plt.pause(0.001)
if trace:
input("Push <ENTER> to finish")
plt.ioff()
return best
def tabu_search(filename, max_iter, tabu_tenure, max_candidates, trace=True):
"""implements the Tabu Search algorithm"""
if trace:
global fig
plt.ion()
fig = plt.figure()
fig.canvas.set_window_title('Tabu Search')
problem = Problem()
problem.load_problem(filename)
#problem.eliminate_time_windows()
start_time = time.time()
current = random_solution(problem)
#current = insertion1_vrptw(problem)
current_cost = current.cost()
if trace:
current.plot()
fig.canvas.draw()
plt.pause(0.0001)
tabu_list = {}
for niter in range(max_iter):
candidates = []
for _ in range(max_candidates):
candidates.append(
generate_candidate(current, list(tabu_list.keys())))
best_candidate, best_edges = min(candidates, key=lambda c: c[0].cost())
best_candidate_cost = best_candidate.cost()
if best_candidate_cost < current_cost:
current = best_candidate
current_cost = best_candidate_cost
if trace:
current.plot()
fig.canvas.draw()
plt.pause(0.0001)
tabu_list[best_edges] = tabu_tenure
for k in list(tabu_list.keys()):
tabu_list[k] -= 1
if tabu_list[k] == 0: del tabu_list[k]
if trace: print(" > iteration=%d, best=%g" % (niter + 1, current_cost))
current.endtime = time.time()
current.starttime = start_time
if trace:
current.plot(True)
plt.pause(0.001)
input("press <ENTER> to finish")
plt.ioff()
return current
def ant_colony_system(filename, max_iter, num_ants, decay, c_hist, c_heur,
c_local_phero, c_greed, trace=True):
"""implements an Ant Colony System for VRPTW"""
if trace:
global fig
plt.ion()
fig = plt.figure()
fig.canvas.set_window_title('Ant Colony System')
problem = Problem()
problem.load_problem(filename)
#problem.eliminate_time_windows()
start_time = time.time()
best = random_solution(problem)
#best = parallel_savings_vrptw(problem)
init_pheromone = 1.0 / ((problem.ncustomers-1) * best.cost())
pheromone = initialise_pheromone_matrix(problem.ncustomers, init_pheromone)
if trace:
best.plot()
fig.canvas.draw()
plt.pause(0.0001)
for i in range(max_iter):
for ant in range(num_ants):
candidate = ant_solution(problem, pheromone, c_hist, c_heur, c_greed)
if trace: print(" ant=%d, cost=%g" % (ant+1, candidate.cost()))
if candidate.cost() < best.cost():
best = candidate
if trace:
best.plot()
fig.canvas.draw()
plt.pause(0.0001)
local_update_pheromone(pheromone, candidate, c_local_phero,
init_pheromone)
global_update_pheromone(pheromone, best, decay)
if trace: print(" > iteration=%d, best=%g" % (i+1, best.cost()))
best.endtime = time.time()
best.starttime = start_time
if trace:
best.plot(True)
plt.pause(0.0001)
input("press <ENTER> to finish")
plt.ioff()
return best
def genetic_algorithm(filename, max_gens, pop_size, p_crossover, p_mutation,
trace=True, histograms=False):
"""implements a Genetic Algorithm for VRPTW"""
if trace:
global fig
plt.ion()
fig = plt.figure(figsize=(14,5.5))
fig.canvas.set_window_title('Genetic Algorithm')
else: histograms = False
problem = Problem()
problem.load_problem(filename)
#problem.eliminate_time_windows()
start_time = time.time()
population = []
# creates the initial population
best = None
for _ in range(pop_size):
individual = random_solution(problem)
if not best or fitness(individual) < fitness(best):
best = copy.copy(individual)
population.append(individual)
# repeats for max_gens generations
perturbative_heuristics()
if trace:
fig
plt.subplot(121)
best.plot()
# trace fitness evolution
gen_update = max(int(max_gens / 100),1)
best_fit, avg_fit, worst_fit = [], [], []
plt.subplot(122)
trace_fitness(population, best_fit, avg_fit, worst_fit, gen_update, pop_size)
fig.canvas.draw()
plt.pause(0.0001)
for gen in range(max_gens):
selected = []
for _ in range(pop_size): selected.append(binary_tournament(population))
children = reproduce(selected, pop_size, p_crossover, p_mutation,
best.cost(), histograms)
children.sort(key=lambda i: fitness(i))
if fitness(children[0]) < fitness(best):
best = copy.copy(children[0])
if trace:
fig
plt.subplot(121)
best.plot()
fig.canvas.draw()
plt.pause(0.0001)
population = children
if trace and gen % gen_update == 0:
# update trace of fitness evolution
fig
plt.subplot(122)
trace_fitness(population, best_fit, avg_fit, worst_fit,
gen_update, pop_size)
fig.canvas.draw()
plt.pause(0.0001)
if trace: print(" > gen %d, best: %g" % (gen+1, best.cost()))
best.endtime = time.time()
best.starttime = start_time
if trace:
fig
plt.subplot(121)
best.plot(True)
plt.pause(0.001)
perturbative_statistics()
if histograms:
PerturbativeHeuristic.plot()
plt.pause(0.001)
if trace:
input("press <ENTER> to finish")
plt.ioff()
return best